| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Systematics and Phytogeography |
2Section of Integrative Biology and Institute of Cellular and Molecular Biology, University of Texas at Austin, 1 University Station A6700, Austin, Texas 78712 USA; 3Department of Botany, Charles Darwin Research Station, Puerto Ayora, Isla Santa Cruz, Galápagos, Ecuador
Received for publication January 20, 2006. Accepted for publication May 11, 2006.
ABSTRACT
Plant biogeographers have long argued whether plant disjunctions result from vicariance or dispersal. One of the classic patterns of plant disjunction involves New World amphitropical disjuncts, as exemplified by Tiquilia subg. Tiquilia (Boraginaceae). Subgenus Tiquilia forms a heterogeneous group of ~20 species that is amphitropically distributed in the deserts of North and South America, with four taxa endemic to the Galápagos Islands. The current study reconstructs the biogeographic history of subg. Tiquilia in order to explore the origins of New World amphitropical disjunction and of Galápagos endemism. A strongly supported phylogeny of the subgenus is estimated using sequence data from matK, ndhF, rps16, ITS, and waxy. Biogeographic analyses using combined and individual marker data sets reveal a complex history of long-distance dispersal in subg. Tiquilia. Biogeographic reconstructions imply a North American origin of the subgenus and its three major lineages and require at least four long-distance dispersal events to explain its current distribution. The South American taxa of subg. Tiquilia result from three independent and nonsimultaneous colonization events, while the monophyly and continental origins of the Galápagos endemics are unresolved. This study contributes to a growing body of evidence that intercontinental dispersal has been more common than previously realized.
Key Words: amphitropical disjunction • Boraginaceae • Galápagos Islands • long-distance dispersal • molecular phylogeny • North America • South America • Tiquilia
Plant biogeographers have long attempted to understand how the numerous plant groups whose geographic ranges are interrupted by thousands of kilometers of open ocean attained their current fragmented distributions (de Queiroz, 2005
). Without the knowledge of continental drift, most early biogeographers hypothesized that transoceanic dispersal was the historical mechanism behind these disjunct ranges (Morrone and Crisci, 1995
; de Queiroz, 2005
). However, the acceptance of plate tectonics in the 1960s and 1970s caused a dramatic shift toward the idea that most plant disjunctions resulted from the fragmentation of earlier, larger landmasses, such as Gondwana (Nelson and Platnick, 1981
; Wiley, 1988
). These vicariance explanations remained dominant until the recent advent of molecular systematic techniques, particularly molecular-based dating of lineage divergences (de Queiroz, 2005
). Using these techniques, much recent scholarship has demonstrated that numerous plant disjunctions are far too young to have resulted from vicariance, leaving transoceanic dispersal as the only plausible alternative (Lavin et al., 2004
; Sanmartín and Ronquist, 2004
; de Queiroz, 2005
; Renner, 2005
). The realization that long-distance dispersal may have been far more frequent than previously supposed has led plant biogeographers using modern molecular tools to reexamine the relative importance of vicariance and dispersal in explaining the classic patterns of worldwide plant disjunction (Givnish and Renner, 2004
; Renner, 2005
).
One of these classic floristic disjunctions involves taxa that are distributed in the boreal, temperate, Mediterranean, or desert regions of both North and South America, but are absent from the intervening tropics (Raven, 1963
, 1972
; Thorne, 1972
). Researchers have debated the causes of this amphitropical disjunction for over a century (Bray, 1898
, 1900
; Johnston, 1940
; Constance, 1963
; Raven, 1963
; Turner, 1972
; Raven and Axelrod, 1974
; Allred, 1981
; Carlquist, 1983
), with disjuncts of the more arid biomes provoking the most discussion historically (Bray, 1898
; Johnston, 1940
; Hunziker et al., 1972
; Solbrig, 1972
; Hunziker, 1975
). To explain this pattern of arid amphitropical distribution, many earlier researchers favored either a migratory hypothesis, whereby ancestral populations dispersed short distances through the tropics across arid or semi-arid "islands" or migrated directly through an ancient, continuous arid to semi-arid tropical corridor along the Pacific coast of the Americas (Raven, 1963
; Solbrig, 1972
; Williams, 1975
), or a hypothesis of parallel evolution of near-identical arid-adapted taxa from widely distributed tropical ancestors (Johnston, 1940
; Barbour, 1969
). More critical studies beginning in the 1960s have led later researchers to favor long-distance dispersal, probably by migrating birds, as the main cause for the disjunctions (Raven, 1963
, 1972
; Cruden, 1966
; Carlquist, 1983
; Simpson and Neff, 1985
).
The amphitropically disjunct plants of North and South America fall into two general phylogenetic classes. The first and larger class is composed of identical or closely related disjunct species or species pairs (see Raven, 1963
, for a thorough list). The morphological similarity exhibited within these species-level disjuncts suggests that they represent very recent dispersal events. Several workers have recently examined these species/species pairs with DNA or protein techniques and have confirmed their phylogenetic closeness and likely recent trans-tropical colonization (examples include Wallace and Jansen, 1990
; Peterson and Ortíz-Díaz, 1998
; Lia et al., 2001
; Soltis et al., 2001
; Beardsley and Olmstead, 2002
; Bleeker et al., 2002
; Li et al., 2002
; Lee et al., 2003
; Beier et al., 2004
). The second class of amphitropical disjuncts comprises disjunct plant groups with multiple species endemic to each continent. Most of these groups occur in the more arid regions of the Americas, with good examples including Ephedra L. (Gnetales; Ickert-Bond and Wojciechowski, 2004
; Huang et al., 2005
), Hoffmannseggia Cav. (Fabaceae; Simpson et al., 2005
), Lycium L./Grabowskia Schlechtd. (Solanaceae; Miller, 2002
; Levin and Miller, 2005
), Menodora Humb. & Bonpl. (Oleaceae; Steyermark, 1932
; Turner, 1991
), and Tiquilia Pers. (Boraginaceae; Richardson, 1977
). These disjunct groups could provide evidence both of earlier dispersal events, which could be inferred in a phylogenetic tree by the presence of a dispersal-derived monophyletic clade of species, and of multiple dispersal events. Under a chance/long-distance dispersal hypothesis, we would expect to find that the dispersal events responsible for arid amphitropical disjunctions have occurred throughout the period of time in which arid and semiarid environments have been available to receive transtropical migrants (since the late Miocene, ~7–5 million years ago; Axelrod, 1979
; Graham, 1999
; Hartley and Chong, 2002
). Therefore, evidence of multiple, non-contemporaneous dispersal events within an amphitropically disjunct group would lend support to long-distance dispersal as the mechanism of arid amphitropical disjunction. The few recent molecular studies of amphitropically disjunct species groups (examples include Lycium/Grabowskia, Miller, 2002
; Levin and Miller, 2005
; Ephedra, Ickert-Bond and Wojciechowski, 2004
; Huang et al., 2005
; and Hoffmannseggia, Simpson et al., 2005
) have suggested that multiple dispersal events may be common in these groups and that such events have not occurred simultaneously.
The genus Tiquilia Pers. (Boraginaceae: Ehretioideae) provides a good opportunity to explore North and South American amphitropical disjunction because it forms a well-defined monophyletic group of arid-adapted plants with multiple species endemic to each continent (Richardson, 1977
; Moore and Jansen, in press
). Tiquilia comprises approximately 30 species of annual to perennial prostrate herbs and subshrubs that are restricted to the arid temperate and subtropical regions of North and South America and has been well monographed by Richardson (1977)
. As part of a study to examine the age and origins of adaptation to aridity in Tiquilia, Moore and Jansen (in press
) found strong molecular support for Richardson's (1977) division of the genus into two monophyletic subgenera, subg. Eddya and subg. Tiquilia. These subgenera are clearly differentiated by morphology, chromosome number, and habitat. They also differ distributionally: subg. Eddya is nearly entirely restricted to the Chihuahuan Desert region of North America, while subg. Tiquilia is distributed amphitropically in the deserts of North (4 spp.) and South (11 spp.) America, as well as in the Galápagos Islands (4 spp.; Fig. 1). One of the species in subg. Tiquilia, T. nuttallii, is itself amphitropically distributed between the Great Basin of western North America and the Monte region of Argentina, while the ~11 South American endemic species are restricted to the deserts of the Peruvian and Chilean coasts. The remaining three North American species form a morphologically heterogeneous assemblage and are found in the Mojave and Sonoran Deserts (Richardson, 1977
).
|
MATERIALS AND METHODS
Taxon and marker selection
All species of subg. Tiquilia except for T. simulans and T. hunteri (which were not encountered in field work) were sequenced for the chloroplast rps16 intron (Oxelman et al., 1997
; Kelchner, 2002
), the nuclear ribosomal internal transcribed spacer (ITS; Baldwin et al., 1995
; Álvarez and Wendel, 2003
) region, and part of the nuclear granule-bound starch synthase gene (GBSSI, or waxy; Mason-Gamer et al., 1998
). Multiple accessions of all North American and some South American/Galápagos species were included (see Appendix for a list of accessions). In most cases, sequences were collected for all three markers for each accession, with the exceptions of some Galápagos accessions (in which amplification of some markers was unsuccessful), T. paronychioides, and most of the other South American mainland taxa (waxy sequences were polymorphic; see Results). The presence or absence of sequence data for a particular marker is indicated for each accession in the Appendix. Preliminary phylogenetic analyses of sequences from these three markers (Moore and Jansen, in press
) allowed the selection of a representative subset of species from subg. Tiquilia for sequencing two additional markers, the chloroplast genes ndhF (Olmstead and Sweere, 1994
) and matK (including the 3' portion of the trnK intron; Johnson and Soltis, 1994
; Steele and Vilgalys, 1994
; Kelchner, 2002
). Sequence data for all five markers were also generated for representative taxa from subg. Eddya and for three outgroup taxa in the Boraginales: Coldenia procumbens, Ehretia anacua, and Bourreria succulenta (Moore and Jansen, in press
).
DNA extraction, amplification, sequencing, and alignment
All accessions of Tiquilia from North and South America were field-collected in silica gel (vouchers deposited at TEX), while accessions from the Galápagos Islands were either field-collected in silica gel or derived from herbarium material (see Appendix). DNA isolation, amplification, sequencing, and alignment utilized the primers and protocols outlined in Moore and Jansen (in press
). Several ITS and waxy sequences for subg. Tiquilia showed evidence of polymorphism, and therefore exemplar accessions for several taxa were cloned for one or both markers (these are noted in the Appendix). Cloning followed the protocols in Moore and Jansen (in press
). All sequences have been deposited in GenBank (accession numbers given in the Appendix), and the final alignments are available in TreeBASE (http://www.treeBASE.org; accession number S1529).
Phylogenetic analyses
A preliminary incongruence length difference (ILD; Farris et al., 1994
) test of combined rps16, ITS, and waxy sequence data for all accessions indicated incongruence among the three marker partitions (P = 0.01) in subg. Tiquilia. The taxa causing the incongruence between the two nuclear data sets were identified as those having multiple non-monophyletic groups of cloned sequences (see Results), while these same taxa in addition to the differing placement of Tiquilia paronychioides between the chloroplast (cpDNA) and nuclear (nrDNA) data sets caused the incongruence between rps16 and the nuclear markers. Because we did not wish to remove this information from the analyses, we chose to analyze each of these three marker sets independently using all available sequences for subg. Tiquilia. The ILD tests involving accessions for which sequence data for all five markers were present also indicated incongruence among three partitions of the data: cpDNA, ITS, and waxy (P = 0.01). Because the incongruence in ITS and waxy involved the position of taxa within the major lineages of subg. Eddya (removing these taxa results in a nonsignificant ILD test result; P = 0.24) rather than in subg. Tiquilia, these markers were combined into a nuclear DNA (nrDNA) data set for a parametric bootstrap analysis (see Tests of alternate topologies). All five markers were congruent (ILD test, P = 0.15) when reduced to 14 ingroup taxa, representing exemplar species for every major lineage in Tiquilia plus two outgroup taxa (Coldenia and Bourreria; Moore and Jansen, in press
). This data set will be referred to as the five-marker combined data set.
Several short regions of ITS (amounting to 13.7% of the total ITS alignment) that were difficult to align among the major lineages of Tiquilia and/or the outgroups were eliminated from the combined analyses. However, ITS was alignable in these regions for two taxon partitions within subg. Tiquilia: for T. plicata + T. cuspidata + T. paronychioides + Galápagos spp. (henceforth called the T. plicata clade), and for T. nuttallii + T. palmeri + ‘blue-flowered' South American spp. (the latter two groups will be referred to as the T. palmeri clade). Because much of the interspecific variation within ITS lay within these highly variable regions, these two partitions were analyzed separately for this marker. These two analyses were largely concerned with the relative positions of T. cuspidata, T. paronychioides, and the Galápagos species for the first partition and T. palmeri and the South American species for the second partition. Because of this and because all preliminary analyses agreed that T. plicata and T. nuttallii, respectively, were clearly located outside of the other taxa in each partition, we used T. plicata to root the first ITS taxon partition and T. nuttallii to root the second partition.
Parsimony and Bayesian searches were conducted for the following data sets: rps16, ITS (split into two analyses, as outlined previously), waxy, combined cpDNA, combined nrDNA, and the five-marker combined data set. Maximum likelihood (ML) searches were also conducted for the three combined data sets, as well as for a reduced version of the waxy data set (see Tests of alternate topologies). The single-marker analyses included all available sequences for subg. Tiquilia (but see below for waxy), while the cpDNA (28 taxa), nrDNA (27 taxa), and five-marker (16 taxa) data sets used representative species for all major lineages. To reduce the search time of the waxy analyses, only one or two sequences were included for each clearly monophyletic group of cloned sequences within each accession. The rps16 and waxy analyses were rooted with Bourreria alone, while the ITS analyses were rooted as described previously. The cpDNA and nrDNA combined data sets included all three outgroups.
Parsminony and ML searches were conducted in PAUP*, version 4.0b10 (Swofford, 2002
) using the parameters described in Moore and Jansen (in press
). Gaps were included in all parsimony analyses, but were coded separately using the simple gap-coding method of Simmons and Ochoterena (2000)
. Clade support was assessed using nonparametric bootstrap analyses using one of two methods. For most data sets, bootstrap analyses involved 100 replicates with the MulTrees setting turned on; however, because of the large number of equally parsimonious trees in the ITS and waxy analyses, 10 000 bootstrap replicates were performed with multrees off for these data sets (Mort et al., 2000
). Bayesian analyses were conducted using MrBayes, version 3.0b4 (Huelsenbeck and Ronquist, 2001
) as described in Moore and Jansen (in press
). The model of sequence evolution for ML and Bayesian analyses was estimated for each data set using the Akaike information criterion (Posada and Buckley, 2004
) as implemented in the program Modeltest, version 3.6 (Posada and Crandall, 1998
).
Tests of alternate topologies
Parametric bootstrapping (also known as the Swofford-Olsen-Waddell-Hillis [SOWH] test; Huelsenbeck et al., 1996
; Swofford et al., 1996
; Goldman et al., 2000
) was employed in two cases to determine whether alternate, suboptimal topologies could be rejected using a given data set. In the first case, a possible incongruence between nuclear and chloroplast data involving the phylogenetic position of T. paronychioides was detected in preliminary analyses (see Results): nrDNA prefer a topology of (T. cuspidata, [T. paronychioides, Galápagos spp.]), while cpDNA prefer a topology of (Galápagos spp., [T. paronychioides, T. cuspidata]). To test this incongruence statistically, a pair of reciprocal bootstrap analyses were conducted in which the nrDNA tree was constrained to have the cpDNA topology with respect to T. paronychioides, and vice versa. In the other case, parametric bootstrapping was conducted using a reduced version of the waxy data set in order to test the monophyly of the T. paronychioides cloned sequences (see Results). For this analysis, the four cloned sequences of T. paronychioides were constrained to be monophyletic. All waxy sequences of subg. Tiquilia that were included in parsimony and Bayesian analyses were included in this parametric bootstrap, except that T. plicata and T. nuttallii were reduced to a single accession in each species. Likewise, only one of the five accessions of subg. Eddya (T. hispidissima 154) was included in the parametric bootstrap.
For all parametric bootstrap analyses, the optimal ML topologies derived from the pertinent original data sets were used to create alternate constraint topologies. New ML searches were then conducted using the topological constraints along with the original data and model parameters for each data set involved. The resulting trees (with branch lengths) and estimated model parameters were used to simulate 500 replicate data sets for each bootstrap analysis in the program Mesquite, version 1.05 (Maddison and Maddison, 2004
). Parsimony searches of these data sets generated a distribution of tree length differences (TLD) between constrained and unconstrained trees. The statistical significance of the observed TLD for each data set was then determined by comparing it to the corresponding simulated distribution in TLD. We chose to implement the parsimony criterion rather than ML in searches of simulated data because of the huge time penalty incurred by ML and because the most parsimonious trees shared identical topologies for Tiquilia with the ML tree for each data set.
Biogeographic analyses
Distribution of species in Tiquilia was included in each data set as a separate character with three possible character states: North America, South America, and Galápagos Islands. Using MacClade, version 4.06 (Maddison and Maddison, 2003
), this character was then optimized using both ACCTRAN and DELTRAN options onto the trees obtained from searches of each data set. Outgroup distribution was not included in biogeographic analyses, because the two closest outgroups consist of monophyletic clades of species with pantropical distributions (including North America) that have no effect on the biogeographic analyses. Dispersal–vicariance (DIVA) analyses (Ronquist, 1996
, 1997
) were also conducted to determine the roles of vicariance and dispersal in the biogeographic history of subg. Tiquilia, with North America, South America, and the Galápagos Islands as characters and taxon presence or absence in each of these localities as character states. Default program parameters were used in all DIVA analyses. For DIVA analyses involving individual marker data sets, polytomies were collapsed to a single exemplar accession/sequence.
RESULTS
Characteristics of each data set are reported in Table 1, along with the model of sequence evolution chosen by Modeltest for each data set. The 5.8S, ndhF, and matK genes had the least amount of variation, while the coding regions of waxy and the noncoding trnK and rps16 introns had an intermediate amount of variation. ITS-1, ITS-2, and the two introns of waxy were highly variable, with regions of ITS unalignable, even in some cases between major lineages of Tiquilia.
|
|
|
|
|
). All but one of these species possesses blue flowers (hence these 10 species will be referred to as the blue-flowered group), and the majority also have exserted stamens and stigmas. In contrast, all other species of Tiquilia are characterized by pale pink or white flowers with inserted stamens and stigmas. The blue-flowered South American species are also all apparently tetraploids, with chromosome numbers (n = 16, with some aneuploid species of 15 and 14) twice that of the North American members of subg. Tiquilia (n = 8; Richardson, 1977
). Bootstrap and Bayesian support for this group was strong in ITS (Fig. 5), but weak for rps16 data (Fig. 4). However, the blue-flowered species T. conspicua and T. elongata were strongly supported as a monophyletic sister group to T. palmeri in the combined cpDNA tree (Fig. 3). Very little divergence was detected among the sequences of the blue-flowered species in any marker. There was little significant support for any subclade involving this group in either ITS or rps16, and the South American species for which multiple accessions were included were not recovered as monophyletic. No evidence of highly divergent types was seen in ITS for these tetraploid species. However, all accessions of these species showed highly similar waxy sequences with evidence of the same divergent types in directly sequenced PCR product. For this reason, we cloned waxy for two of the accessions (T. elongata 289 and T. conspicua 294), representing species with divergent morphology. The cloned sequences were similar for each accession and fell into two strongly supported monophyletic groups: one group formed a well-supported monophyletic clade with the Colorado River group of T. palmeri, while the other group was either weakly supported as sister to the Salton Sea group of T. palmeri (58% bootstrap proportion in the parsimony analysis, with separate gap characters included) or as basal to the T. palmeri clade (0.85 posterior probability in the Bayesian analysis, with gap characters removed; Fig. 7). The position of the latter group of sequences in the Bayesian analysis agreed with the position of the blue-flowered group in all other analyses.
|
|
Biogeographic analyses
DIVA analyses and biogeographic reconstruction using MacClade were congruent across all data sets and indicated that Tiquilia, both subgenera, and all three major lineages of subg. Tiquilia originated in North America (Figs. 2, 3). DIVA analyses required four or five dispersal events to explain the current distribution of subg. Tiquilia, including at least three dispersals to the South American mainland and at least one to the Galápagos Islands. Two of the dispersal events, involving the colonization of the Monte region of Argentina by T. nuttallii and the colonization of the coastal desert of South America by the ancestor of the blue-flowered clade, clearly proceeded directly from North America to South America. The remaining two or three dispersal events involved the T. plicata clade. The sequence of events in the biogeographic history of this clade was unresolved because of the topological incongruence among the different markers for this group. The cpDNA tree indicated two dispersals from North America: one directly to South America, giving rise to T. paronychioides, and one directly to the Galápagos Islands, giving rise to the island clade (Fig. 3). The individual nuclear marker analyses resulted in a complicated and sometimes conflicting set of biogeographic reconstructions. The ITS analyses required only two dispersals to explain the distribution of the T. plicata clade, with an initial dispersal from North America. However, biogeographic reconstructions were unable to resolve whether this initial colonization event proceeded to South America or the Galápagos Islands. At least one subsequent dispersal was required by biogeographic reconstructions using the ITS trees, either from South America to the Galápagos or vice versa, depending on the direction of the initial dispersal event. Assuming that T. paronychioides is an allotetraploid (see Discussion), MPTs from waxy indicated a suite of possible dispersal histories for the T. plicata clade, involving two or three dispersal events. All of the differing reconstructions implied for the T. plicata clade by the various marker analyses are illustrated in simplified form in Fig. 8.
|
Origin and dispersal history of subg. Tiquilia
All phylogenetic analyses indicate that Tiquilia, its subgenera, and all major lineages of subg. Tiquilia originated in North America, a result consistent with the North American concentration of morphological variation within the genus. All phylogenetic analyses also agree that at least one long-distance dispersal event to South America has occurred in each of the major lineages of subg. Tiquilia (T. nuttallii, the T. palmeri clade, and the T. plicata clade; Figs. 2–7). The different colonization histories of the three major lineages are discussed below.
Dispersal in Tiquilia nuttallii
The biogeographic reconstructions suggest that T. nuttallii is a species of North American origin that has relatively recently colonized the deserts of Argentina. The North and South American populations of this species are morphologically indistinguishable, and likewise, there is no meaningful sequence variation among North and South American populations in any marker (Figs. 4, 5, 7). Tiquilia nuttallii is both the northernmost and southernmost species in the genus, and so its dispersal to South America probably represents the longest-distance dispersal in the history of the genus. This dispersal event may have been facilitated by its life history traits and likely breeding system. Tiquilia nuttallii is the only true annual in the genus and therefore produces a large number of fruits every year. It is also likely a selfer, judging from its minute, inconspicuous flowers, which are easily the smallest among the North American species (Richardson, 1977
). All of these characteristics would tend to increase the likelihood of establishment in a new suitable locality, and indeed, these same traits have been observed in many of the amphitropically distributed herbaceous plants of North and South America (Raven, 1963
).
Dispersal in the Tiquilia palmeri clade
The monophyly of the blue-flowered group of South American coastal desert species suggests that these species result from a single dispersal event to South America by a relative of the Sonoran/Mojave Desert T. palmeri. Although a broad range of morphological novelty has evolved among the blue-flowered group, there is very little phylogenetically informative sequence diversity among these species (Figs. 4, 5). This lack of molecular diversity points to a relatively rapid morphological radiation in this clade. Judging by the great variation in flower size, color, and structure among these species (Richardson, 1977
), this radiation seems to have been driven at least partly by adaptation to new pollinators. Although the pollinators of the blue-flowered South American species are unknown, North American members of subg. Tiquilia are typically pollinated by a number of small Tiquilia-specializing bees, including species in subgenera Heteroperdita and Epimacrotera of the large panurgine genus Perdita (J. L. Neff, Central Texas Melittological Institute, Austin, Texas, USA, personal communication). These bees are absent in South America, so the earliest ancestor of the blue-flowered group that dispersed to South America would have been under strong selection pressure to adapt to new pollinators. Such adaptive radiations following chance long-distance dispersal are common in oceanic island taxa (Francisco-Ortega et al., 1996
; Crawford and Stuessy, 1997
; Baldwin, 1998
; Crawford et al.,
1998; Givnish, 1998
; Kim et al., 1999
; Lindqvist and Albert, 2002
). Although the coastal deserts of Perú and Chile are continental regions, they do form an arid habitat "island" on which the blue-flowered clade has apparently radiated, possibly in response to a combination of factors including not only pollinator differences but also geographic separation of ancestral populations among the various coastal lomas and higher-elevation desert communities (Richardson, 1977
; Rundel et al., 1991
).
The waxy sequence data suggest that the blue-flowered species are allotetraploids and clearly identify one of the two diploid parents. Allotetraploid taxa would be expected to have duplicate waxy loci, inherited from each parent, which would explain the two monophyletic groups of waxy sequences detected in the two blue-flowered species that were cloned (Tiquilia conspicua and T. elongata; Fig. 7). The Colorado River group of T. palmeri likely provided one of the two diploid parents of the blue-flowered clade, judging from the strongly supported monophyly of these T. palmeri and one of the two blue-flowered clone groups. However, the differing positions of the other blue-flowered cloned sequences in the parsimony and Bayesian analyses of waxy parsimony prohibit the precise identification of the second diploid parent of the blue-flowered species. Either this second parent was a member of the Salton Sea group of T. palmeri (based on parsimony, with gaps included) or belonged to a now-extinct or hitherto unsampled lineage of the T. palmeri clade (based on Bayesian analyses, with gaps excluded). Unlike waxy, the ITS data do not demonstrate evidence of multiple types in the ITS data for the blue-flowered group. ITS is known to evolve by concerted evolution (Álvarez and Wendel, 2003
), and it is possible that one of the putative diploid parental types could have been eliminated as a result of this process.
Dispersal in the Tiquilia plicata clade
Biogeographic analyses of the rps16, ITS, and waxy trees generated a confusing set of conflicting dispersal histories in the T. plicata clade (depicted in simplified form in Fig. 8). All analyses require at least two long-distance dispersal events in this clade and agree that at least one of these dispersals originated in North America, but the different markers otherwise disagree on the directions and number of dispersals implied for this clade.
The independent origins of South American T. paronychioides and the Galápagos Islands Tiquilia implied by the cpDNA analyses (Figs. 3, 4) seem unsatisfactory in light of the great morphological similarity between the Galápagos endemics and South American T. paronychioides (Richardson, 1977
). Using the cpDNA topology, such similarity would necessarily be interpreted either as the result of convergent evolution or of extinction in North America; in other words, the extinct North American ancestor of this clade must have been similar to the Galápagos Islands taxa and T. paronychioides, while the extant North American species T. cuspidata must represent morphological novelty. In contrast, the nuclear markers support a close relationship between the Galápagos taxa and T. paronychioides. These two lineages form a strongly supported monophyletic group in both ITS (Fig. 6) and waxy (Fig. 7), which is fully congruent with the morphological evidence. Although the three Galápagos species T. darwinii, T. fusca, and T. galapagoa are supported as monophyletic in both nuclear marker analyses, the positions of the Galápagos endemic T. nesiotica and the South American T. paronychioides are unclear with respect to the remaining Galápagos taxa. This phylogenetic ambiguity results in a number of competing biogeographic reconstructions in both the ITS and waxy analyses. The implications of these differing scenarios are discussed briefly.
The ITS data suggest three possible histories of dispersal in the T. plicata clade, all involving two dispersal events (illustrated in Fig. 8B). The first of these possible scenarios, in which the Galápagos taxa are derived from South American ancestors, is congruent with recent work suggesting that the majority of the Galápagos endemic flora is of South American origin (Porter, 1983
; Eliasson, 1990
; Elisens, 1992
; Schilling et al., 1994
; Peralta and Spooner, 2001
; Walsh and Hoot, 2001
) and might therefore seem most reasonable. However, a North American origin has been convincingly demonstrated for the Galápagos endemic Gossypium klotzschianum Andersson (Wendel and Percival, 1990
) and has been postulated for several others (Porter, 1983
). Moreover, the nearest continental relative of G. klotzschianum occurs in Baja California as does T. cuspidata, the nearest relative of T. paronychioides and the Galápagos Tiquilia. Likewise, although it may seem less likely that an island endemic group would spawn a continental derivative rather than vice versa, such island-to-continent dispersals have been suggested in several Macaronesian plant groups (Moore et al., 2002
; Mort et al., 2002
; Carine et al., 2004
). It would therefore be premature to rule out either of the remaining two ITS-based dispersal scenarios.
The cloned waxy sequences for T. paronychioides further complicate the dispersal history of the T. plicata clade. As in the allotetraploid blue-flowered clade, the waxy sequences of T. paronychioides are resolved in two different positions (Fig. 7). Richardson (1977)
obtained an approximate polyploid chromosome count for this species, and the sister relationship of the Galápagos clone group of T. paronychioides with the Galápagos Tiquilia, while not conclusive, is suggestive of an allotetraploid origin for this species. This hypothesis is further supported by the fact that all uncloned accessions of T. paronychioides displayed clear evidence of possessing both waxy clone groups in directly sequenced PCR product. However, the parametric bootstrapping analysis of the waxy data does not rule out the monophyly of the T. paronychioides cloned sequences, and so the possibility of diploidy in this species remains. In contrast, the Galápagos species of subg. Tiquilia appear to be diploid. Although no karyological information exists for these taxa, no evidence of multiple waxy types was present in any of the three Galápagos accessions.
Assuming that T. paronychioides and the Galápagos taxa are respectively allotetraploid and diploid, two possible dispersal scenarios can account for their current distributions. The first requires that the Galápagos waxy clones of T. paronychioides be derived from Galápagos ancestors. Under such a hypothesis, biogeographic analyses using the waxy topology indicate that three equally parsimonious histories of dispersal can account for the distribution of the T. plicata clade (Fig. 8C). All three reconstructions require three dispersal events, as well as at least one dispersal event from the Galápagos Islands to South America, followed by an allotetraploid event to form T. paronychioides.
An alternative and perhaps more satisfying general scenario for the distribution of waxy types in T. paronychioides assumes that the Galápagos taxa are derived from one of two continental diploid ancestors of T. paronychioides. Under this hypothesis, the Galápagos sequence type of T. paronychioides represents a potentially extinct South American ancestor of both allotetraploid T. paronychioides and the Galápagos taxa. Assuming this phylogenetic history, the Galápagos taxa are of South American origin and result from one or two dispersal events, depending on which most parsimonious waxy topology of the Galápagos taxa is used (Fig. 8D). This general scenario, which assumes that the putative allotetraploid event leading to T. paronychioides occurred in South America after the colonization of the Galápagos Islands, seems more likely than the previous Galápagos-to-South America scenario because it can explain the distribution of the T. plicata clade with two rather than three dispersal events. Furthermore, allotetraploidy resulting from hybridization of two closely related continental species seems more plausible than hybridization of a continental species with a migrant from a distant archipelago. Still, we cannot rule out either scenario given the current data.
Of the entire set of phylogeographic hypotheses outlined above, the differing phylogeographic histories implied by the nrDNA and cpDNA data are most difficult to reconcile. While differences among the individual nuclear marker biogeographic reconstructions result largely from topological uncertainty, parametric bootstrapping indicates that the topological incongruence in the T. plicata clade between combined cpDNA and nrDNA data sets is statistically significant and therefore is likely the result of differing genome histories. Such real phylogenetic incongruence is often explained as the result of introgression or hybridization (Wendel and Doyle, 1998
), but such an explanation in the T. plicata clade is complicated by the great geographic separation of the North American, South American, and Galápagos taxa. Thus any contact among these lineages subsequent to the assumption of their current distributions could have only occurred through long-distance dispersal. A number of explanations could account for the observed conflict in cpDNA and nrDNA, including ancient hybridization or lineage sorting (Wendel and Doyle, 1998
), but all are impossible to prove with the current data.
Timing of dispersal events in subg. Tiquilia
Although we cannot precisely determine the dates at which the various long-distance dispersal events occurred, it seems highly likely from the phylogenetic analyses that they were not simultaneous. Of the three major lineages in subg. Tiquilia, the most recent dispersal from North to South America is almost certainly in T. nuttallii. The sequence uniformity across all markers in North and South American populations of this species, in conjunction with the the restricted range of T. nuttallii in Argentina, strongly suggests a very recent migration to South America, possibly only in the last few thousand years. In contrast, the North-to-South dispersal events in the remaining two lineages would appear to have occurred much earlier. Both the T. palmeri and T. plicata clades are composed of multiple taxa in South America and/or the Galápagos Islands, and unlike in T. nuttallii, these Southern Hemisphere clades are morphologically distinct from their Northern Hemisphere relatives. While the morphological diversity present in the southern members of both of these clades appears to have arisen relatively quickly, such diversity still requires a reasonable period of time to evolve. Judging by the arid sites to which both of these clades are restricted at present, it is unlikely that either clade could have existed in South America or the Galápagos Islands prior to the first appearance of aridity in these regions approximately 6 mya and possibly not before the onset of permanent aridity around 3 mya (Hartley and Chong, 2002
). A dated ndhF phylogeny of the Boraginales including exemplar taxa of all major lineages of Tiquilia (Moore and Jansen, in press
) agrees with such a biogeographic scenario. This dated phylogeny indicates that neither the blue-flowered species nor the Galápagos taxa/T. paronychiodes likely diverged prior to ~6 mya. Assuming a lag between the first appearance of the North American ancestor of each of these groups and their dispersal southward, it is reasonable to suppose that both of these groups colonized South America and the Galápagos archipelago not long after the beginning of permanent aridity in these regions ~3 mya.
Conclusions
Phylogenetic analyses indicate that subg. Tiquilia and its three major lineages originated in North America and that at least four long-distance dispersals are required to explain the current distribution of the subgenus in North America, South America, and the Galápagos Islands. Biogeographic reconstruction also indicates that the South American mainland taxa are derived from at least three independent colonization events: one event gave rise to Argentine T. nuttallii, one event in the T. palmeri clade gave rise to the blue-flowered South American group, while a further event in the T. plicata clade gave rise to South American T. paronychioides. The dispersal history of the T. plicata clade is unresolved, with cpDNA, ITS, and waxy data suggesting different and sometimes competing hypotheses. The Galápagos taxa may result from one or two introductions to the archipelago. The origin of this group is placed in North America by cpDNA, but nuclear data suggest either a North or South American origin. The dispersal history between the Galápagos Islands and the South American mainland in the T. plicata clade is also unclear. Neither nuclear marker is able to resolve whether T. paronychioides colonized the Galápagos Islands or is an island-derived species. The waxy data further confuse the situation by suggesting that T. paronychioides is an allotetraploid, with one parent possibly from the Galápagos Islands.
All available evidence points to repeated long-distance dispersal as the mechanism leading to the amphitropical disjunction in subg. Tiquilia. Morphological and molecular evidence suggests that the three introductions of subg. Tiquilia to South America almost certainly occurred separately rather than simultaneously, as would be expected under a chance dispersal hypothesis. Furthermore, molecular dating indicates that these introductions have occurred in the last few million years, during a time when there is no evidence of a continuous arid or semi-arid corridor suitable for an arid-adapted genus like Tiquilia (Burnham and Graham, 1999
). The impressive floral radiation of the blue-flowered clade of South American species also supports long-distance dispersal as the mechanism of disjunction. The possible adaptation to new pollinators suggested by this floral diversity is precisely what would be expected in an outcrossing group that was removed from its normal pollinating fauna by chance long-distance dispersal.
The present work corroborates other recent molecular studies that have found repeated long-distance dispersal in amphitropically disjunct plant groups. In addition to Tiquilia, the amphitropical distributions of three other genera have now been inferred to result from multiple dispersals: Osmorhiza Raf. (Apiaceae; three dispersals, probably all North to South America; Wen et al., 2002
), Sanicula L. (Apiaceae; two dispersals, both North to South America; Vargas et al., 1998
), and Hoffmannseggia (Fabaceae; four dispersals, all South to North America; Simpson et al., 2005
). The multiple dispersal events documented in Tiquilia and other similar disjunct groups, when taken together with the numerous examples of recent long-distance dispersal in species-level amphitropical disjuncts, imply that long-distance dispersal between North and South America has been relatively common during the history of the arid, temperate, and boreal floras of the Americas. This observation adds weight to the growing consensus that intercontinental dispersal has been far more important in the assembly of modern floras worldwide than previously recognized (Lavin et al., 2004
; de Queiroz, 2005
; Renner, 2005
).
For taxa/accessions with asterisks: *Taxon/Accession—GenBank accessions: ITS, matK, ndhF, rps16, waxy; Collection info; Voucher specimen.
For taxa/accessions without asterisks: Taxon/Accession—GenBank accessions: ITS, rps16, waxy; Source; Voucher specimen .
Outgroups
*Bourreria succulenta Jacq.—DQ197285, DQ197229, DQ197257, DQ197037, DQ197599; Cuba: Pinar del Rio; R. G. Olmstead 96–114 (WTU). *Coldenia procumbens L.—DQ197284, DQ197227, DQ197255, DQ197036, DQ197597; Ghana: Bolgatanga; Jongkind & Nieuwenhuis 1973 (MO). *Ehretia anacua I.M. Johnst.—DQ197286, DQ197228, DQ197256, DQ197038, DQ197600; Texas: Travis County, cult. at University of Texas at Austin; M. J. Moore s.n.
Tiquilia Pers.
SubgenusEddya A.T. Richardson
*Tiquilia canescens (DC.) A.T. Richardson var.canescens—DQ197312, DQ197230, DQ197258, DQ197068, DQ197630; Nevada: Clark County, Sandy Valley Road at Columbia Pass; M. J. Moore 239. *T."Durango" (undescribed species)—DQ197331, DQ197232, DQ197260, DQ197141, DQ197649; Durango: MEX 30, 40 mi W of Mapimí; M. J. Moore 260. T. gossypina (Wooton and Standl.) A.T. Richardson: *T. gossypina.134—DQ197337, DQ197233, DQ197267, DQ197146, DQ197653; Texas: Brewster County, FM 170, 9 mi W of Terlingua. *T. gossypina.263—DQ197353, DQ197234, DQ197263, DQ197158, DQ197667; Coahuila: 52 mi N of S end of Coahuila 91, Valle de Acatita. *T. greggii (Torr. & A. Gray) A.T. Richardson—DQ197325, DQ197231, DQ197259, DQ197083, DQ197644; Texas: Brewster County, Study Butte; M. J. Moore 133. T. hispidissima (Torr. & A. Gray) A.T. Richardson: *T. hispidissima.131—DQ197423, DQ197240, DQ197268, DQ197098, DQ197727; Texas: Brewster County, Study Butte. *T. hispidissima.154—DQ197442, DQ197241, DQ197269, DQ197101, DQ197734; Texas: Culberson County, TX 54, 31.5 mi N of Van Horn. *T. hispidissima.185—DQ197485, DQ197242, DQ197270, DQ197121, DQ197763; New Mexico: Socorro County, US 380, 21 mi W of Carrizozo. T. latior (I.M. Johnst.) A.T. Richardson: *T. latior.211—DQ197535, DQ197243, DQ197271, DQ197136, DQ197777; Arizona: Navajo County, AZ 77, 5.5 mi S of Holbrook. *T. latior.216—DQ197538, DQ197244, DQ197272, DQ197139, DQ197786; Utah: Wayne County, UT 24, 6 mi N of Hanksville. T. mexicana (S. Watson) A.T. Richardson: *T. mexicana.117—DQ197368, DQ197235, DQ197261, DQ197167, DQ197682; Texas: Terrell County, US 90, 10 mi E of Dryden. *T. mexicana.245—DQ197372, DQ197236, DQ197266, DQ197172, DQ197692; Chihuahua, MEX 18, 11 mi S of Ojinaga. *T. purpusii (Brandegee) A.T. Richardson—DQ197409, DQ197245, DQ197273, DQ197090, DQ197789; San Luís Potosí: MEX 80, 1 km E of Presa de Guadalupe; M. J. Moore 109. T. tuberculata A.T. Richardson: *T. tuberculata.93C—DQ197402, DQ197238, DQ197264, DQ197184, DQ197711; Nuevo León: MEX 53, near KM 101. *T. tuberculata.98—DQ197407, DQ197239, DQ197265, DQ197186, DQ197717; Nuevo León: near Espinazo. *T. turneri A.T. Richardson—DQ197398, DQ197237, DQ197262, DQ197181, DQ197706; Coahuila: Bolsón de Cuatro Ciénegas; M. J. Moore 89.
SubgenusTiquilia
Tiquilia atacamensis (Phil.) A.T. Richardson: Tatacam.8447—DQ197585, DQ197214, —; Chile: KM 57–58 on road from Calama to San Pedro de Atacama; J. L. Panero and B. Crozier 8447A. T. conspicua (I.M. Johnst.) A.T. Richardson: *Tcons.294—DQ197586, DQ197250, DQ197278, DQ197216, (C1 DQ197816, *C4 DQ197817); Perú: Dpto. Arequipa, Mollendo. Tcfcons.297— —, DQ197215, —; Perú: Dpto. Arequipa, Pan American Hwy, near KM 850. T. cuspidata (I.M. Johnst.) A.T. Richardson: *Tcusp.223—DQ197540, DQ197247, DQ197277, DQ197192, DQ197795; Baja California Sur: MEX 1 near KM 7. Tcusp.224— —, DQ197193, DQ197796; Baja California Sur, MEX 1 near KM 127. T. darwinii (Hook. f.) A.T. Richardson: *Tdar.573—DQ197541, DQ197248, DQ197276, DQ197194, DQ197797; Galápagos Islands: Isla Santiago, Playa Montañitas de Arena; A. Tye 573 (CDS). Tdar.635—DQ197542, DQ197195, DQ197798; Galápagos Islands: Isla Santiago, Sullivan Bay; A. Tye 635 (CDS). Tdar.2252—DQ197543, —, —; Galápagos Islands: Isla Santiago; A. P. Yánez 2252. T. dichotoma (Ruiz & Pavon) Pers.: Tdich.304—DQ197587, DQ197217, —; Perú: Dpto. La Libertad, Pan American Hwy between KM 545 and 550. T. elongata (Rusby) A.T. Richardson: *Telong.289—DQ197588, DQ197251, DQ197279, DQ197218, (C1 DQ197818, *C4 DQ197819); Perú: Dpto. Arequipa, KM 14 of road from Arequipa to Yura. Telong.290—DQ197589, DQ197219, —; Perú: Dpto. Arequipa, KM 14 of road from Arequipa to Yura. T. ferreyrae (I.M. Johnst.) A.T. Richardson: Tferr.296—DQ197590, DQ197220, —; Perú: Dpto. Arequipa, Pan American Hwy, at turnoff to Quilca. Tferr.301—DQ197591, DQ197221, —; Perú: Dpto. Arequipa: Pan American Hwy, near KM 520. T. fusca (Hook.f) A.T. Richardson: Tfus.597—DQ197544, DQ197196, —; Galápagos Islands: Isla San Cristóbal, Bahía Rosa Blanca; A. Tye 597 (CDS). Tfus.598—DQ197545, —, —; Galápagos Islands: Isla San Cristóbal, Sappho Cove; A. Tye 598 (CDS). T. galapagoa (J.T. Howell) A.T. Richardson: Tgalap.120—DQ197546, —, —; Galápagos Islands: Isla Santa Cruz, Cerro Dragón: A. Tye 120 (CDS; from herbarium material). Tgalap.1053—(C1 DQ197547, C2 DQ197548, C3 DQ197549, C4 DQ197550, C5 DQ197551, C7 DQ197552, C9 DQ197553, C10 DQ197554, C11 DQ197555), —, —; Galápagos Islands: Isla Santa Cruz, "La Torta"; I. Aldaz 1053 (CDS; from herbarium material). T. grandiflora (Phil.) A.T. Richardson: Tgrand.293—DQ197592, DQ197222, —; Perú: Dpto. Arequipa, KM 16 along the road from Arequipa to Cerro Verde. T. litoralis (Phil.) A.T. Richardson: Tlit.295—DQ197593, DQ197223, —; Perú: Dpto. Arequipa, Pan American Hwy, at turnoff to Quilca. Tlit.299—DQ197594, DQ197224, —; Perú: Dpto. Arequipa: Pan American Hwy, near KM 625. T. nesiotica (J.T. Howell) A.T. Richardson: Tnes.658—(C1 DQ197556, C3 DQ197557, C4 DQ197558, C5 DQ197559, C6 DQ197560, C8 DQ197561, C10 DQ197562), DQ197197, DQ197799; Galápagos Islands: Isla Bartolomé; A. Tye 658 (CDS). T. nuttallii (Benth.) A.T. Richardson: Tnutt.53—DQ197575, DQ197187, DQ197811; Argentina: Prov. Mendoza, near El Sosneado. Tnutt.55—DQ197576, DQ197188, DQ197812; Argentina: Prov. Mendoza, near Malarguë. Tnutt.65—DQ197577, —, —; Argentina: Prov. Mendoza, near El Manzano. Tnutt.207—DQ197578, DQ197189, DQ197813; Utah: Washington County, jct of UT 219 and road to Enterprise Reservoir. *Tnutt.218—DQ197579, DQ197254, DQ197282, DQ197190, DQ197814; Washington: Grant County, CR 2 SW, 0.2 mi W of jct with CR K SW. Tnutt.1491—DQ197580, DQ197191, DQ197815; Nevada: Churchill County, 3 mi N of Lyon County line on NV 95; B. E. Goodson 1491. T. palmeri (A. Gr
